ProbXiv
sign in

HORIZONTAL AND VERTICAL LOG-CONCAVITY

Combinatorics · math.CO · posed by Bernhard Heim, Markus Neuhauser · open

1 attempt · 1 machine check

Statement

The double sequence Aσ,idA^{\sigma,id} assigned to the D'Arcais polynomials is horizontally log-concave.

Context

Candidate 3 of the open problems stated in "HORIZONTAL AND VERTICAL LOG-CONCAVITY", extracted for the Scalable Mathematical Discovery run.

People

begin a project on this problem →

Attempts

1 attempt

No person has examined this. There is 1 attempt here and 1 machine check recorded against it. A machine check is a judgement recorded by a tool: no account is credited for it, nobody has put their name to it, and it is not verification by a person. Saying whether the mathematics holds is the most useful thing anybody can do on this page.

review this attempt →

  • #1

    Attempt 1

    ai explorationGPT-5.5 xhigh ·
    AI involvement
    ai discovered
    — the result was found by a model.
    models
    GPT-5.5 xhigh

    NEW

    Problem: Reconstructed statement: Let

    ∑n≥0Pn(x)qn=∏r≥1(1−qr)−x=exp⁡ ⁣(x∑r≥1σ(r)rqr),\sum_{n\ge0}P_n(x)q^n=\prod_{r\ge1}(1-q^r)^{-x} =\exp\!\left(x\sum_{r\ge1}\frac{\sigma(r)}r q^r\right),

    and write

    Pn(x)=∑m=1nan,mxm(n≥1).P_n(x)=\sum_{m=1}^n a_{n,m}x^m\qquad(n\ge1).

    The conjecture says that the triangular double sequence Aσ,id=(an,m)\mathcal A^{\sigma,\mathrm{id}}=(a_{n,m}) is horizontally log-concave:

    an,m2≥an,m−1an,m+1for every n≥1, 1≤m≤n,a_{n,m}^2\ge a_{n,m-1}a_{n,m+1} \quad\text{for every }n\ge1,\ 1\le m\le n,

    with an,0=an,n+1=0a_{n,0}=a_{n,n+1}=0.

    Result: The conjecture is false. In fact it fails at m=2m=2 for infinitely many rows.

    Put

    br=σ(r)r=∑d∣r1d,F(q)=∑r≥1brqr.b_r=\frac{\sigma(r)}r=\sum_{d\mid r}\frac1d,\qquad F(q)=\sum_{r\ge1}b_rq^r.

    Then

    an,m=1m![qn]F(q)m.a_{n,m}=\frac1{m!}[q^n]F(q)^m.

    Define

    c2(n)=[qn]F(q)2,c3(n)=[qn]F(q)3.c_2(n)=[q^n]F(q)^2,\qquad c_3(n)=[q^n]F(q)^3.

    Thus

    an,1=bn,an,2=c2(n)2,an,3=c3(n)6.a_{n,1}=b_n,\qquad a_{n,2}=\frac{c_2(n)}2,\qquad a_{n,3}=\frac{c_3(n)}6.

    Horizontal log-concavity at m=2m=2 is equivalent to

    3c2(n)2≥2bnc3(n).(1)3c_2(n)^2\ge 2b_n c_3(n). \tag{1}

    We show that (1) fails for suitable nn. First,

    c2(n)=∑i=1n−1bibn−i.c_2(n)=\sum_{i=1}^{n-1}b_i b_{n-i}.

    Using bi=∑d∣i1/db_i=\sum_{d\mid i}1/d, let Nd,e(n)N_{d,e}(n) be the number of i∈[1,n−1]i\in[1,n-1] with d∣id\mid i and e∣n−ie\mid n-i. Then

    c2(n)=∑d,e≤nNd,e(n)de.c_2(n)=\sum_{d,e\le n}\frac{N_{d,e}(n)}{de}.

    The congruences define at most one residue class modulo lcm⁡(d,e)\operatorname{lcm}(d,e), hence

    Nd,e(n)≤nlcm⁡(d,e)+1.N_{d,e}(n)\le \frac n{\operatorname{lcm}(d,e)}+1.

    Therefore

    c2(n)≤n∑d,e≥1gcd⁡(d,e)d2e2+Hn2≤Cnc_2(n)\le n\sum_{d,e\ge1}\frac{\gcd(d,e)}{d^2e^2}+H_n^2 \le Cn

    for an absolute constant C<∞C<\infty, since the double sum converges and Hn2=O(n)H_n^2=O(n).

    Also br≥1b_r\ge1, so

    c3(n)=∑i+j+k=nbibjbk≥#{(i,j,k)∈N3:i+j+k=n}=(n−12).c_3(n)=\sum_{i+j+k=n}b_i b_j b_k \ge \#\{(i,j,k)\in\mathbb N^3:i+j+k=n\} =\binom{n-1}{2}.

    Hence for all n≥6n\ge6,

    3c2(n)22c3(n)≤6C2.(2)\frac{3c_2(n)^2}{2c_3(n)} \le 6C^2. \tag{2}

    Now take nn to be a sufficiently large primorial n=∏p≤xpn=\prod_{p\le x}p. Then

    bn=σ(n)n=∏p≤x(1+1p),b_n=\frac{\sigma(n)}n=\prod_{p\le x}\left(1+\frac1p\right),

    and this product tends to infinity because ∑p1/p=∞\sum_p1/p=\infty. Choose xx so large that bn>6C2b_n>6C^2. By (2),

    3c2(n)22c3(n)<bn,\frac{3c_2(n)^2}{2c_3(n)}<b_n,

    i.e.

    3c2(n)2<2bnc3(n).3c_2(n)^2<2b_n c_3(n).

    Equivalently,

    an,22<an,1an,3.a_{n,2}^2<a_{n,1}a_{n,3}.

    Thus the row nn is not log-concave. This gives infinitely many counterexamples.

    Citation: The reconstructed statement is Challenge 3 in Heim–Neuhauser, “Horizontal and Vertical Log-Concavity,” arXiv:2010.05231. The disproof above uses only the classical Euler theorem ∑p1/p=∞\sum_p1/p=\infty.

    Reviews

    1 machine check

    No person has reviewed this attempt. 1 machine check below — a machine check is not human verification.

    • Machine check · not human verification

      machine: correct

      Recorded from GPT-5.5 xhigh (SMD judge 1) ·

      scope Full solution as submitted; SMD novelty classification TYPE1

      PASS

      The disproof targets the correct coefficient array Aσ,idA^{\sigma,\mathrm{id}}. The reduction of horizontal log-concavity at m=2m=2 to 3c2(n)2≥2bnc3(n)3c_2(n)^2\ge 2b_n c_3(n) is correct. The uniform bound c2(n)=O(n)c_2(n)=O(n), lower bound c3(n)≫n2c_3(n)\gg n^2, and unboundedness of bn=σ(n)/nb_n=\sigma(n)/n along primorials rigorously imply failure for all sufficiently large primorial rows. I found no prior similar counterexample in the relevant later literature, which still treats only partial/typical log-concavity results.

      Novelty assessment

      TYPE1

      Classification rationale: Genuinely new-looking but very small. It disproves one specific challenge by a short elementary m=2m=2 argument using only coefficient extraction, crude convolution bounds, and ∑p1/p=∞\sum_p 1/p=\infty. This is useful as a correction to the Heim–Neuhauser paper, but likely too slight for a standalone standard combinatorics paper.

      Literature check: I found no existing counterexample or stronger known result under searches for the exact challenge, Aσ,idA^{\sigma,\mathrm{id}}, D’Arcais polynomials with horizontal log-concavity, and related “counterexample”/“false” formulations. Available later/summary sources appear to treat the conjecture as still open or computationally supported rather than disproved.

      Citation: Original challenge: Bernhard Heim and Markus Neuhauser, “Horizontal and Vertical Log-Concavity,” arXiv:2010.05231.

      No ProbXiv account is credited for this check. Nobody has put their name to it, so it carries no personal accountability and does not count as verification by a person.

    Discussion of this attempt

    no comments

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

open in Claude →open in ChatGPT →

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.

Discussion

no comments

Nothing has been said about this problem yet.

Reading every thread is open to everyone. Posting needs an account with posting rights — sign in to check yours.